This invention relates to data communications and more specifically to control of gain, amplitude or threshold in a data communications receiver with soft decision decoding to obtain an optimum signal level into a soft decision decoder.
Data transmission and reception systems that operate on noisy channels use any of a variety of coding and modulation schemes known in the art. Many of these schemes allow a data communications receiver to make symbol decisions directly on the output of a demodulator and feed the decisions to an error detection or correction decoder. Improved performance in such a system is obtained by obtaining soft decision information about the reliability of those received symbols and using that information in the decoder.
If the decoder is implemented in hardware, it is important to conserve logic gates in the decoder by using the minimum acceptable number of bits of precision to represent the soft decision information. It is well known in the art that three-bit quantization produces nearly the same error rates as infinite precision soft decision decoding if the signal and noise amplitudes are appropriate for the quantization thresholds.
When a data communication receiver uses quantized soft decision information with an error correction decoder, some means must be used to set the amplitude of the signal at the quantization point. Too large or too small amplitude does not capture the full benefit from the soft decision information, resulting in more bit errors than the optimum amplitude setting. Amplitude control is typically achieved by automatic gain control circuits that require an amplitude detector or noise level measurement. In many cases, a simpler method of control is desirable.
The prior art generally uses an automatic gain control (AGC) circuit that operates to maintain a constant measure of either signal plus noise or noise alone. An Advanced Hardware Architectures, Inc Application Note ANRS07-0795xe2x80x9cSoft Decision Thresholds and Effects on Viterbi Performancexe2x80x9d indicates that a noise variance AGC is desirable, but that a signal plus noise AGC is easier to implement. A Qualcomm Application Note AN1650-2 xe2x80x9cSetting Soft-Decision Thresholds for Viterbi Decoder Code Words from PSK Modemsxe2x80x9d states that the quantizer thresholds are based on noise alone and that the AGC should operate on noise. Practical AGC algorithms do not operate on noise alone but use signal plus noise or typically on signal alone. The application note illustrates a signal plus noise controlled feedback AGC.
U.S. Pat. No. 5,566,191 discloses a soft decision decoding method using a Viterbi decoder. The need to control the level of an input signal is described and a feedforward AGC rather than the more typical feedback AGC is disclosed. The AGC feeds forward a signal dependant on the amplitude of the received signal to correct soft decision likelihood metric values in the Viterbi decoder.
U.S. Pat. Nos. 5,214,675 and 5,379,324 disclose noise variance estimation for use in receiver that corrects for Rayleigh fading in a multipath channel.
What is needed is a simple gain control function to set the input to a soft decision decoder at a near-optimum level to obtain the best performance in a data communications system.
A data communications receiver incorporating soft decision decoding is disclosed. The data communications receiver includes a demodulator for demodulating an input signal, a means connected to the demodulator for producing strong and weak data symbols from the demodulated input signal and for controlling a desired fraction of strong and weak data symbols, and a soft decision decoder for receiving and decoding the strong and weak data symbols to produce a data output.
In one embodiment of the present invention, the means for producing the strong and weak data symbols from the demodulated signal may be a soft decision quantizer. The means for controlling the desired fraction of strong and weak data symbols to the soft decision decoder comprises a strong/weak indication function connected to the soft decision quantizer for providing strong and weak indications. An averaging function connected to the strong/weak indication function averages the strong/weak indications. A comparison function connected to the averaging function is used to compare the average of strong/weak indications to a desired fraction of weak signals to generate a gain error signal. A gain control function generates a gain control signal from the gain error signal received from the comparison function. A variable gain element connected to the input of the receiver and to the gain control function varies the gain with the gain control signal to maintain the desired fraction of strong and weak data symbols to the soft decision decoder.
In another embodiment of the present invention feed forward gain control is used rather than feedback. The demodulator provides a multiple bit digital word output indicative of the demodulated input signal. The means for producing strong and weak data symbols of the demodulated input signal and for controlling the desired fraction of strong and weak data symbols provided to the decoder comprises strong/weak indication functions, averaging functions, a threshold value, comparison functions, a logic function, and a bit shifter. Each strong/weak indication function has an input connected to a hard decision bit output from the demodulator and another input connected to another output bit of the demodulator. The averaging functions connected to outputs of the strong/weak indication functions average the strong/weak indications. A fixed threshold value is used to set a threshold to control the desired fraction of strong/weak indications. The comparison functions connected to the averaging functions and to the threshold value are used to indicate which strong/weak indications exceed the threshold. A logic function connected to the comparison functions is used to select an index number indicative of which strong/weak indications drop below a threshold and to generate a shift control signal. A bit shifter connected to the multiple bit digital word output of the demodulator and to the logic function receives the shift control signal to select bits from the demodulator that provide the desired fraction of weak and strong symbols to the soft decision decoder. The bit shifter functions as a gain control means.
In another embodiment of the present invention, the threshold is varied rather than the gain. The means for producing strong and weak data symbols of the demodulated input signal and for controlling the desired fraction of strong and weak data symbols further comprises a first greater/than test function, an absolute value circuit, a plurality of greater/than test functions, a threshold control function, and scaling functions. The first greater/than test function has a first input connected to the demodulator to receive the demodulated input signal and a second input connected to a reference to provide a hard decision indication. The absolute value circuit connected to the demodulator receives the demodulated input signal to provide an absolute value output of the demodulated input signal. In the plurality of greater/than test functions, each has a first input connected to the absolute value circuit output. A second greater/than test function in the plurality of greater/than test functions provides a strong/weak indication. The threshold control function is connected to the strong/weak indication to provide a variable threshold output level in accordance with the strong/weak indication. The scaling functions are connected to the variable threshold output and to a second input of the greater/than test functions. The scaling functions scale the thresholds to provide a desired fraction of strong and weak data symbols from the greater/than test functions to the soft decision decoder.
It is an object of the present invention to provide a means to set the amplitude or threshold for quantized soft decisions in a near-optimum manner.
It is an object of the present invention to set the amplitude or threshold for quantized soft decisions with a simple method of control.
It is an advantage of the present invention to measure the fraction of the soft decisions which mark bits as weak or strong and to adjust the gain to achieve a desired fraction of each marking.